Scintillation detectors are widely used in modern physics, including various medical tomographs. There are two types of position-sensitive scintillation detectors: pixel arrays of "scintillator-photomultiplier" pairs and continuous scintillator with multiple photomultipliers attached. For the last one, there are two common ways to calculate the coordinates of the scintillation flash via detector response in scintillation cameras: Anger's method and Monte-Carlo simulation. In this article, we develop an analytical approach (so-called "kaleidoscopic ray-tracing") to compute the energy outcome from several light detectors in the scintillation camera with a continuous scintillator, which has reflective and absorbing faces. We show that Anger's method is just an approximation of the exact solution. Anger's method requires absorbing faces only, so our approach covers a more general case. The derived method achieves results, that are similar to the Monte-Carlo simulation, but it requires fewer computations.